Power Distribution Network (PDN) Impedance and Target Impedance
(Room Ballroom H)
18 Oct 18
2:00 PM
-
2:30 PM
Many people in the industry are somewhat confused about the meaning and definitions of target impedance and actual PDN impedance. This paper will address these issues. Target impedance is a simple Ohms Law calculation. It involves the circuit voltage tolerance, often 5% of Vdd, and the transient current. Z_target = Vdd x tol / I_transient . The transient current is often the most difficult part of this equation and is often hard to find. A good starting point is I_max, the maximum current that the VRM will ever deliver. The target impedance is nearly always a function of frequency because the transient current is a function of frequency. Power gating and clock gating are not used in some products. If a logic clock is always on, the highest transient current is likely to be less than 60% of I_max. The harshest current condition for a PDN is when some portion of the transient current can start and stop at a PDN resonant frequency. Thankfully, this rarely happens and if it does, it is only a small portion of I_max. The actual PDN impedance should not stray very far from the target impedance. A PDN that has impedance significantly below the target impedance is more costly than necessary and a PDN that exceeds the target impedance is in danger of failing. A carefully calculated target impedance is the starting point for good performance and cost optimization for a PDN. Next we take a look at the actual impedance at a frequency point, i.e. 100 mOhms at 100 MHz. This point alone does not tell us very much about how the PDN will perform. The impedance at a frequency may be a point on a long flat line or it may be the top of an impedance peak with a q-factor of 3 or even 10. These 3 PDN impedance profiles perform very differently when stimulated with different waveforms. By examining three specific current waveforms (impulse, step and resonance) a PDN is completely characterized. The three PDN types are simulated with the three fundamental waveforms and graphics are presented. The PDN response to the current waveforms are calculated using closed form equations and demonstrated with simulation. By working backwards from the actual impedance and the target impedance equation, we determine the maximum amount of transient current that a PDN can sustain with acceptable voltage droop or overshoot. This quantifies the relationship between target impedance and actual PDN impedance. Many successful consumer products have actual PDN impedances that greatly exceed target impedance because the probability of pathological current waveforms actually occurring is very low. On the other hand, a PDN that does not exceed the target impedance in any frequency band can exceed the voltage tolerance because of the phenomenon known as rogue waves. PDN design is a matter of managing risk, cost and performance. Target impedance is not a law, it a reference line that is useful for cost-effective and robust PDN design.